Using Machine Learning to Discover Evolutionary Relationships

Abstract

Complex microbial communities exist within and on many organisms, ranging from those that form symbiotic relationships with their host to those whose relationships are seemingly inert.  It has been hypothesized that these microbial communities evolve along with, and may in part drive, the evolution of the host species.  In our study, we use high-throughput DNA sequencing to examine the skin and gut microbiome, both bacterial and fungal, of nine closely related slimy salamanders (Plethodon glutinosus and  P. jordani species complexes), which are terrestrial members of the lungless-salamander family, Plethodontidae.    We characterized the salamanders as a holobiont to determine if the microbial communities correlate with the evolutionary divergence of host nuclear and mitochondrial gene trees.  We hypothesized that the host species in recent stages of speciation have a skin and gut microbiome that is actively diverging and that the host–microbiome relationship is the result of a complex co-evolutionary ⁄ colonization phenomenon.  Nine salamander clades were assessed for differences in microbial assemblage across clade and geographic location.  Our objective is to leverage the analytical capabilities of machine learning to determine if a discernable delineation exists between these clades based upon these microbial differences.